Skip to main content

Table 5 Evaluation of segmentation results with different loss function

From: Automated segmentation and diagnosis of pneumothorax on chest X-rays with fully convolutional multi-scale ScSE-DenseNet: a retrospective study

Method

U-Net [18]

SegNet [32]

DeepLab v3+ [33]

DenseASPP [34]

FC-DenseNet [25]

MS_scSE_FC-DenseNet

\({\mathrm {DSC}}_1\) Mean(Standard Deviation)

 CEL

0.78(0.34)

0.81(0.31)

0.78(0.33)

0.77(0.33)

0.82(0.30)

0.82(0.29)*

 W-CEL

0.79(0.34)

0.80(0.32)

0.77(0.34)

0.76(0.34)*

0.83(0.29)

0.81(0.30)*

 SW-CEL

0.79(0.33)

0.80(0.31)

0.78(0.32)

0.78(0.32)

0.82(0.29)

0.84(0.27)

HD Max(Mean)

 CEL

19.63(3.24)*

17.21(2.53)

17.90(2.80)

20.48(3.20)*

15.47(2.23)

17.81(2.27)*

 W-CEL

19.99(2.80)*

17.21(2.53)

18.41(3.01)

19.74(3.71)

16.22(2.03)*

15.25(2.03)

 SW-CEL

17.76(2.60)

16.34(2.27)

19.56(2.99)

18.52(2.74)

14.81(2.02)

14.87(1.95)